Using MD5 or SHA1 for instance, and applying integers (as seed so to speak) to the hash function, in sequence, and only keeping, say, the first 64 bits of the resulting hash, do we always have a probability close to $1/{2^{64}}$ to have $$\text{hash}_{64}(n) = \text{hash}_{64}(n+1)$$ for every $n\in\big[1,2^{64}]$?